Make your software understand your non-US accent
|Project:||NZE-vox, Pocketsphinx, Julius|
Free speech recognition engines like Sphinx are not hard-wired for any particular language and accent; they use statistical models tuned for the desired dialect. Unfortunately, there are not very many free models of non-American English, let alone other languages.
This talk looks at what is necessary to create free acoustic and language models for New Zealand and Australian English.
Douglas Bagnall is an artist and programmer who is currently working on an exhibit requiring speech recognition. He lives in Wellington, New Zealand.